基于神经网络的聚碳酸酯表面形貌参数与光泽度关系研究

研究了聚碳酸酯复杂表面形貌参数与表面光泽度的关系.使用不同目数的砂纸作为刮头,施加不同大小的恒定法向力对聚碳酸酯高光表面进行刮擦,得到具有不同光泽度的损伤表面.使用原子力显微镜和光泽度仪分别测得材料表面的三维形貌参数和光泽度数据,发现高光聚碳酸酯表面在砂纸的作用下损伤越严重,光泽度值越低.采用皮尔逊相关系数分析单个形貌参数与表面光泽度间的相关性,发现在所研究的16个形貌参数中,只有8个形貌参数与光泽度有较强的相关性.选用BP和RBF神经网络,对具有较强相关性的形貌参数与表面光泽度的复杂关系进行学习和预测,结果显示2种神经网络模型都能够预测表面损伤后光泽度的改变,基于对2种模型的分析,RBF神经网络比BP神经网络更适合于建立形貌参数与光泽度的非线性联系.

Investigation of Correlation Between Surface Topography Parameters and Gloss of Polycarbonate Based on Neural Network

The relationship between the complex surface topography parameters and surface gloss of polycarbonate was systematically studied.The original high gloss polycarbonate plates were scratched by the sandpaper with different grit numbers under different levels of constant normal load.The three-dimensional topography parameters and the gloss level of the sample surfaces were obtained by using atomic force microscopy and gloss meter respectively.The Pearson correlation coefficient was adopted to estimate the correlation between individual topographic parameters and the surface gloss.The complex relationship between the surface gloss and those topographic parameters having strong correlation coefficient were further investigated using BP(Back Propagation)and RBF(Radial Basis Function)neural networks.The results showed that the gloss level decreased with the increasing damage degree for the high gloss polycarbonate plates.The correlation analysis showed that only 8 out of 16 topographic parameters had strong correlation with the surface gloss.While both neural networks are able to predict the change of gloss after surface scratch,RBF network was identified to be more suitable than BP network for establishing the nonlinear relationship between the topography parameters and the surface gloss.